论文部分内容阅读
Lasso二元选择分位数回归是较为新颖的统计方法,一方面通过Lasso的变量选择功能,能够从众多的影响因素中识别出关键因素;另一方面通过分位数回归对各因素在不同分位点处的异质影响进行细致刻画,能够获得更多信息进而实现信用状况的准确评估,可望在信用评估领域发挥重要作用。基于Lasso二元选择分位数回归,建立评估模型并将其应用于中国上市公司的信用评估。通过数值模拟和实证研究,将其与基于Logit回归、Lasso-Logit回归和支持向量机的评估效果进行对比,发现前者不但具备良好的变量选择能力而且可以获得最佳的评估效果。
Lasso binary choice quantile regression is a relatively new statistical method. On the one hand, Lasso’s variable selection function can identify the key factors from many influencing factors; on the other hand, by quantile regression, Site of the heterogeneous impact of a detailed description of the ability to get more information to achieve an accurate credit rating assessment is expected to play an important role in the field of credit evaluation. Based on Lasso binary choice quantile regression, an evaluation model was established and applied to the credit rating of Chinese listed companies. Through numerical simulation and empirical study, it is compared with the evaluation results based on Logit regression, Lasso-Logit regression and support vector machines, and found that the former not only has good ability of variable selection but also can obtain the best assessment results.